Go to https://machinelearningforkids.co.uk/ and click the Get Started button you then want to Try without registering. Then press Add New Project.
Name your project: “Kiwi, Kiwi, or Kiwi?”
Choose an Recognise Image project (because we’re training from pictures).
Choose Storage “In your web browser”
Press CREATE
Click on your new project
Create two classes to start by clicking the Add new Label button:
Kiwi bird
Kiwifruit
Find 5 pictures of Kiwi birds and 5 pictures of Kiwifruit and put them into your project. One easy way to do this is open two browsers and drag them from one into your project. If you get the images from google then you need to drag them from the enlarged image after you click on the image not the cached version.
When you have added your pictures click the < Back to Project button at the top of the page.
Next press Learn and Test, then Train new machine learning model.
You are now going to test your model. The easiest way is to right click on the enlarged image in google images and Copy Image Address and Paste that into the box. Test three kiwi bird images and three kiwifruit. The images cannot be images that you used in your training data.
You are now going to add more data so you have 10 pictures of both. Retrain your model with the new data and try testing it again.
Find some images that are still of kiwis but might trip up your models. Examples could include: Try a kiwi bird toy, kiwi fruit sticker, or drawing—does it still work?
Add a third label to your model and call it Kiwi Humans. Include 10 pictures of famous New Zealanders and Train your model again. This time test your model with just two examples of each (kiwi bird, kiwifruit, kiwi human).
Think about and you may want do some reaserch about the three refection questions as a class.